Merge pull request #124 from jmejia8/add-rdex-algorithm

Add RDEx algorithm

Author: jmejia8

.github/workflows/CI.yml

@@ -13,7 +13,7 @@ jobs:
       fail-fast: false
       matrix:
         version:
-          - 1.7
+          - 1.10.10
         os:
           - ubuntu-latest
         arch:

Project.toml

@@ -1,10 +1,11 @@
 name = "Metaheuristics"
 uuid = "bcdb8e00-2c21-11e9-3065-2b553b22f898"
-authors = ["Jesus Mejia <jesusmejded@gmail.com>"]
 version = "3.4.2"
+authors = ["Jesus Mejia <jesusmejded@gmail.com>"]
 
 [deps]
 Distances = "b4f34e82-e78d-54a5-968a-f98e89d6e8f7"
+Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
 JMcDM = "358108f5-d052-4d0a-8344-d5384e00c0e5"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
@@ -18,19 +19,19 @@ Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 [compat]
 Aqua = "^0.8"
 Distances = "^0.10.7"
+Distributions = "^0.25"
 JMcDM = "^0.7.1"
 LinearAlgebra = "1"
 Printf = "1"
 Random = "1"
 Reexport = "^1"
 Requires = "1"
-Statistics = "1"
 SearchSpaces = "^0.2"
 SnoopPrecompile = "1"
+Statistics = "1"
 Test = "1"
-UnicodePlots = "2.6, 3"
-
-julia = "^1.7"
+UnicodePlots = "^2.6, ^3"
+julia = "^1.10"
 
 [extras]
 Aqua = "4c88cf16-eb10-579e-8560-4a9242c79595"

src/Metaheuristics.jl

@@ -30,6 +30,7 @@ export optimize!
 export set_user_solutions!
 export boxconstraints
 export SHADE, CSO
+export RDEx
 
 include("externals.jl")
 

src/algorithms/singleobjective/RDEx/RDEx.jl

@@ -0,0 +1,35 @@
+"""
+RDEx: An effectiveness-driven hybrid Single-Objective optimizer
+
+Julia implementation of the adaptive differential evolution algorithm.
+
+"""
+module RDExModule
+
+# Standard library dependencies
+using Random
+using Statistics
+using Distributions
+
+# Submodules
+include("RDEx/constants.jl")
+include("RDEx/random.jl")
+include("RDEx/selection.jl")
+include("RDEx/types.jl")
+include("RDEx/mutation.jl")
+include("RDEx/adaptation.jl")
+include("RDEx/metaheuristics.jl")
+
+
+# Export public API
+export RDEx
+
+
+
+# Import Metaheuristics API (avoid conflict by importing functions directly)
+import .MetaheuristicsAPI: RDEx 
+# Note: RDExAlgorithm is the Metaheuristics-compatible algorithm constructor
+# Users can do: using RDEx: RDExAlgorithm; algo = RDExAlgorithm()
+# Or: import RDEx: RDExAlgorithm as RDEx; algo = RDEx()
+
+end # module RDExModule

src/algorithms/singleobjective/RDEx/RDEx/adaptation.jl

@@ -0,0 +1,111 @@
+"""
+Parameter adaptation mechanisms for RDEx optimizer.
+
+Handles adaptation of F, Cr, and hybrid strategy parameters.
+"""
+module AdaptationUtils
+
+import ..Constants
+
+"""
+    weighted_mean(values::Vector{Float64}, weights::Vector{Float64}, success_filled::Int, weights_storage::Vector{Float64}) -> Float64
+
+Compute weighted mean of values using weights.
+"""
+function weighted_mean(
+    values::Vector{Float64},
+    weights::Vector{Float64},
+    success_filled::Int,
+    weights_storage::Vector{Float64}
+)::Float64
+    sum_weight = sum(weights[1:success_filled])
+    if sum_weight == 0.0
+        return 1.0
+    end
+    
+    weights_storage[1:success_filled] = weights[1:success_filled] / sum_weight
+    
+    sum_square = sum(weights_storage[i] * values[i] * values[i] for i = 1:success_filled)
+    sum_val = sum(weights_storage[i] * values[i] for i = 1:success_filled)
+    
+    if abs(sum_val) > 1e-8
+        return sum_square / sum_val
+    else
+        return 1.0
+    end
+end
+
+"""
+    update_memory!(
+        parameters,
+        temp_success_cr::Vector{Float64},
+        temp_success_f::Vector{Float64},
+        fit_delta::Vector{Float64}
+    )
+
+Update memory for Cr and F parameters based on successful trials.
+"""
+function update_memory!(
+    parameters,
+    temp_success_cr::Vector{Float64},
+    temp_success_f::Vector{Float64},
+    fit_delta::Vector{Float64}
+)::Nothing
+    if parameters.success_filled != 0
+        parameters.memory_cr[parameters.memory_iter+1] = 0.5 * (
+            weighted_mean(temp_success_cr, fit_delta, parameters.success_filled, parameters.weights) + 
+            parameters.memory_cr[parameters.memory_iter+1]
+        )
+        parameters.memory_f[parameters.memory_iter+1] = weighted_mean(
+            temp_success_f, fit_delta, parameters.success_filled, parameters.weights
+        )
+        parameters.memory_iter = (parameters.memory_iter + 1) % parameters.memory_size
+    end
+    return nothing
+end
+
+"""
+    update_eb_hybrid_param!(
+        parameters,
+        eb_hybrid_flag::Vector{Float64},
+        fit_arr_front::Vector{Float64},
+        fit_temp::Vector{Float64},
+        n_inds_front::Int
+    )
+
+Update the elite-based hybrid rate parameter based on performance.
+"""
+function update_eb_hybrid_param!(
+    parameters,
+    eb_hybrid_flag::Vector{Float64},
+    fit_arr_front::Vector{Float64},
+    fit_temp::Vector{Float64},
+    n_inds_front::Int
+)::Nothing
+    sum_eb_delta_fit = 0.0
+    sum_origin_delta_fit = 0.0
+    
+    for chosen_one = 1:n_inds_front
+        if eb_hybrid_flag[chosen_one] == 1.0
+            if fit_temp[chosen_one] <= fit_arr_front[chosen_one]
+                sum_eb_delta_fit += fit_arr_front[chosen_one] - fit_temp[chosen_one]
+            end
+        else
+            if fit_temp[chosen_one] <= fit_arr_front[chosen_one]
+                sum_origin_delta_fit += fit_arr_front[chosen_one] - fit_temp[chosen_one]
+            end
+        end
+    end
+    
+    if sum_eb_delta_fit != 0.0 && sum_origin_delta_fit != 0.0
+        parameters.eb_hybrid_rate = sum_eb_delta_fit / (sum_eb_delta_fit + sum_origin_delta_fit)
+        parameters.eb_hybrid_rate = clamp(parameters.eb_hybrid_rate, 0.0, 1.0)
+    else
+        parameters.eb_hybrid_rate = Constants.EB_HYBRID_RATE_INIT
+    end
+    
+    return nothing
+end
+
+end # module AdaptationUtils
+

src/algorithms/singleobjective/RDEx/RDEx/constants.jl

@@ -0,0 +1,11 @@
+"""
+Constants used throughout the RDEx optimizer.
+"""
+module Constants
+
+# Algorithm constants
+const EB_HYBRID_RATE_INIT = 0.7
+const MEMORY_SIZE = 5
+
+end # module Constants
+